At present, it is quite common to find people totally newbye in CFD trying to use code for doing also very complex simulations. And, is also quite common to see any kind of questions and doubts that can be easily solved with a minimum knowledge of the basic CFD (just have a look to the questions at the cfd-online forum).

There are several levels to learn and to do CFD, each one requiring a mix between the knowledge of the basis of CFD and the knowledge of some CFD code.

Thinking to learn just how to set the GUI without any knowledge of CFD is dangeorus. A CFD software is not user-friendly even if a GUI seems it is. And running a CFD code is not like using a remote TV command.

That does not mean to be necessarily an expert CFD software developer but my warning is to be, at least, a smart CFD user.

It is important if you want to understand why you get the results you get. It is even more important if you want to understand why you do not get the results you would like to get.

Hi Ramadan Bassiouny,

as I can say from my little experience and with regard to generating CFD results for real problems/processes and not for just studying the numerical theory itself:

It seems impossible to get started with CFD and to know all the possible important details from the beginning. And even more, if you are a long time into the field CFD you will probably not know everything there is to know, since its a very broad and multi discipline field.

Coding is a good way to get started to get the basics, but its unrealistically to develop a full featured CFD code on your own, so at some point your learning curve relating to pure CFD knowledge will probably slow down a lot, due to problems with handling the code's complexity. Theory on its own on the other hand is very abstract and complicated, especially if you are going from the differential equations to their numerical solution and the algebra theory of the numerical solution of large linear systems. Also you may have bugs or compiler dependencies in your software, which may influence your solution and you are not really aware of anyhow (as you may read here: Article Reproducible and replicable CFD: it's harder than you think

So to conclude with, in the best case you should probably have a good imagination about the things you are going to simulate (you have to know your problem/process) and also have a proper knowledge of CFD either to generate some usable results. Trying to simulate problem related things, were you may have access to validation data, is a good way to get there and I think that's why its done in so many research projects at universities.

At least to add some of my personal two cents: I think today its generally better to use opensource (CFD) codes like OpenFOAM over commercial ones, at least for educational purposes or for the application in public funded projects, for the following reasons:

• They are often very mature in their basis
• The user is forced into deeper knowledge acquisition
• The user is more independent of features added by the software producers
• Opensource code makes the reproducibility a lot better than commercial codes, where even significant bugs or changes of certain versions may not be published (which is understandable by the way, from a companies point of view...)
• Users can more easily become developers
• The general benefit for the public is greater
• Self educational step for the usage of commercial products for graduate students is a lot smaller than the other way around

Best regards,

Christian

At present, it is quite common to find people totally newbye in CFD trying to use code for doing also very complex simulations. And, is also quite common to see any kind of questions and doubts that can be easily solved with a minimum knowledge of the basic CFD (just have a look to the questions at the cfd-online forum).

There are several levels to learn and to do CFD, each one requiring a mix between the knowledge of the basis of CFD and the knowledge of some CFD code.

Thinking to learn just how to set the GUI without any knowledge of CFD is dangeorus. A CFD software is not user-friendly even if a GUI seems it is. And running a CFD code is not like using a remote TV command.

That does not mean to be necessarily an expert CFD software developer but my warning is to be, at least, a smart CFD user.

Hi,

Yes it is necessary to have solid knowledge of CFD methodology to obtain good results .

How to assign boundary conditions, choosing schemes either first order or second order ,turbulence models based on the problem etc all are vital to obtain and understand CFD.

Without proper knowledge of the methodology , the package would turn out to be Confusing Fluid Dynamics .

it will make one understand ,what exactly is CFD commercial code...that is very important for coming up with boundry conditions other than default..

You can obtain the results, but it's unrealistic to get the concise and reasonable results at your first try! Otherwise, you cannot improve your model or implement modifications without having the basic knowledge on the internal codes and numerical schemes. Furthermore, in most cases you need to extend the basic code in some parts for your own objectives and that is also not possible without understanding the solver's methodology.

Using the CFD software without knowing the numerical method, boundary conditions along with the source terms if any, and the stability criterion applicable for a particular problem in study will result in wrong outcome. Therefore, the CFD software will be like a black box without the knowledge of numerical methods.

Of course you should have some experiences with the relevant basics about numerical methods.However, the most proper thing is to have the talent to read from time to time and discuss numerical issues with expertise to get the clue of each part of these packages.

Recently, CFD vendors tend to sell CFD software to people who don't have CFD basics. It may be possible to simulate single phase flow, but as for multi-phase flow calculation, CFD basics are required for properly simulations.

Unfortunately I see also some courses at University that are like the user guide for the software...

I firmly believe the most vital thing is to have a very good understanding of fluid dynamics and the phenomena you are trying to model. Without that, there is a danger that the model extent and/or the boundary conditions and/or the mesh will be inappropriate. And to know where to look in the post-processing to pick out the key features of interest and to check that the model is behaving as you expect. The next most important thing is then to understand the turbulence modelling. And that's just single phase where the physics is well understood!

It is of course important to understand the basis behind CFD modelling (e.g. VOF) and the limitations imposed, but not necessarily the fine detail of the numerical solutions (unless that's the subject of your research!).

Its common to see students esp. at undergrad levels falling prey to using CFD packages with superficial to no knowledge of numerical methods at work in the background. While its important to understand the numerical methodology underlying the package, one can hardly argue that people are not getting it right without any knowledge of such concepts. In my humble opinion, its a good thing to know these concepts but not knowing wont get you trapped anywhere.

I think, like other researchers who posted respons above that, for correctly application of CFD packages it is very important to have a solid basic training in the field of fluid engineering, fluid mechanics, applied mathematics (including numerical methods) and in the field of fluid engineering.

That, because CFD software is only a step in the chain of steps to be taken to simulate of fluid engineering systems.

The first and a very important step is to build a conceptual model which requires among other the Definition and the space and time limitation of the modelled system, which is indeed a sub-system (i.e. a cut) of a general system. The interaction between both leads to concrete boundary conditions for the modelled sub-system. To a correct solving of this step good basics knowledge in the field of fluid engineering are required.

To describe the flow processes, we use mathematical equations i.e. the Navier-Stokes Equation which is the governing equation of CFD. Consequently, the next step is the mathematical-physical problem formulation (e.g. initial and boundary value problem formulation). Further to solve the formulated initial/boundary/value problem by computer, we must translate it to the discretized form. The translators are numerical discretization methods, such as Finite Difference, Finite Element, Finite Volume methods and boundary element method (only the boundary discretization required). These steps require good basics knowledge in the field of applied mathematics.

At the end, we must analyse the simulation results comparing with experiments and the real problem. If the results are not enough to solve the problem, we must repeat the simulation process until find satisfied solution. This end step like the first step required also good basics knowledge in the field of fluid engineering.

We should compare the cfd results with water models wherever possible. Very complex problems can be solved on cfd packages like Ansys and fluent. I personally believe numerical methods can be used for finding out initial conditions like velocity at inlet, temperature at some wall. After putting right boundary conditions and initial conditions, fluent does the rest for you.

Amiy Srivastava

"After putting right boundary conditions and initial conditions, fluent does the rest for you"

If you think this statement is the truth I think you will be disappointed in your career ...

Oh my God, seriously! That's why we do validate our results with water modeling. Can you throw some light on that statement related to disappointment.

There is no sense in writing "...put the right BC...fluent does the rest for you".

After you will discover why during your career we will discuss about the issues.

"Water modeling" means nothing.

An old joke says that numerical people are used to go to experimental people to have comfort in their measurements and experimental people used to go to numerical people to have comfort in numerical simulations.

Ok..Got it. Thanks a lot